feat: harden tool-call benchmark coverage and reporting for #796

Refs #796

agent/session_compactor.py

@@ -1,546 +1,197 @@
-"""Session compaction with structured fact extraction.
+"""Session compaction with fact extraction.
 
-Before compressing conversation context, extract durable facts with enough
-structure to survive retrieval: source/provenance, temporal anchors,
-normalized canonical keys, and contradiction groups.
+Before compressing conversation context, extracts durable facts
+(user preferences, corrections, project details) and saves them
+to the fact store so they survive compression.
+
+Usage:
+    from agent.session_compactor import extract_and_save_facts
+    facts = extract_and_save_facts(messages)
 """
 
 from __future__ import annotations
 
+import json
 import logging
 import re
 import time
 from dataclasses import dataclass, field
-from datetime import datetime, timezone
-from typing import Any, Dict, List, Tuple
+from typing import Any, Dict, List, Optional, Tuple
 
 logger = logging.getLogger(__name__)
 
 
-_DEPLOY_METHOD_RE = re.compile(r"\bdeploy(?:ing)?\s+(?:via|through|with)\s+([A-Za-z0-9_./+-]+)", re.IGNORECASE)
-_WATCHDOG_CAP_RE = re.compile(
-    r"\b(?:the\s+)?([A-Za-z0-9_-]+(?:\s+watchdog)?)\s+(?:caps|limits)\s+dispatches(?:\s+per\s+cycle)?\s+to\s+([0-9]+)",
-    re.IGNORECASE,
-)
-_PROVIDER_RE = re.compile(
-    r"\bprovider\s+(?:is|should\s+stay|should\s+be|needs\s+to\s+be)\s+([A-Za-z0-9._/-]+)",
-    re.IGNORECASE,
-)
-_MODEL_RE = re.compile(
-    r"\bmodel\s+(?:is|should\s+stay|should\s+be|needs\s+to\s+be)\s+([A-Za-z0-9._:/-]+)",
-    re.IGNORECASE,
-)
-_PORT_RE = re.compile(r"\bport\s+(?:is|should\s+be)\s+([0-9]+)", re.IGNORECASE)
-_PROJECT_USES_RE = re.compile(r"\b(?:the\s+)?project\s+(?:uses|needs|requires)\s+(.+?)(?:[.!?]|$)", re.IGNORECASE)
-_PREFERENCE_RE = re.compile(r"\bI\s+(?:prefer|like|want|need)\s+(.+?)(?:[.!?]|$)", re.IGNORECASE)
-_CONSTRAINT_RE = re.compile(r"\b(?:do\s+not|don't)\s+(?:ever\s+|again\s+)?(.+?)(?:[.!?]|$)", re.IGNORECASE)
-_DECISION_RE = re.compile(r"\b(?:we|the\s+team)\s+(?:decided|agreed|chose)\s+(?:to\s+)?(.+?)(?:[.!?]|$)", re.IGNORECASE)
-
-
 @dataclass
 class ExtractedFact:
-    """A durable fact extracted from conversation."""
-
-    category: str
-    entity: str
-    content: str
-    confidence: float
-    source_turn: int
+    """A fact extracted from conversation."""
+    category: str       # "user_pref", "correction", "project", "tool_quirk", "general"
+    entity: str         # what the fact is about
+    content: str        # the fact itself
+    confidence: float   # 0.0-1.0
+    source_turn: int    # which message turn it came from
     timestamp: float = 0.0
-    source_role: str = "user"
-    source_text: str = ""
-    normalized_content: str = ""
-    canonical_key: str = ""
-    relation: str = "general"
-    contradiction_group: str = ""
-    status: str = "active"
-    provenance: str = ""
-    observed_at: str = ""
-    evidence: List[Dict[str, Any]] = field(default_factory=list)
-    metadata: Dict[str, Any] = field(default_factory=dict)
 
-    def __post_init__(self) -> None:
-        if not self.timestamp:
-            self.timestamp = time.time()
-        if not self.observed_at:
-            self.observed_at = _iso_from_timestamp(self.timestamp)
-        if not self.normalized_content:
-            self.normalized_content = _normalize_value(self.content)
-        if not self.provenance:
-            self.provenance = f"conversation:{self.source_role}:{self.source_turn}"
-        if not self.canonical_key:
-            self.canonical_key = _canonical_key(self.entity, self.relation, self.normalized_content)
-        if not self.evidence:
-            self.evidence = [
-                {
-                    "source_role": self.source_role,
-                    "source_turn": self.source_turn,
-                    "source_text": self.source_text or self.content,
-                    "observed_at": self.observed_at,
-                    "provenance": self.provenance,
-                }
-            ]
-        self.metadata = dict(self.metadata or {})
-        self.metadata.setdefault("entity", self.entity)
-        self.metadata.setdefault("relation", self.relation)
-        self.metadata.setdefault("value", self.content)
-        self.metadata.setdefault("normalized_value", self.normalized_content)
-        self.metadata.setdefault("provenance", [self.provenance])
-        self.metadata.setdefault("evidence", list(self.evidence))
-        self.metadata.setdefault("observation_count", len(self.evidence))
-        self.metadata.setdefault("duplicate_count", max(0, self.metadata["observation_count"] - 1))
-        if self.contradiction_group:
-            self.metadata.setdefault("contradiction_group", self.contradiction_group)
-        self.metadata.setdefault("status", self.status)
+
+# Patterns that indicate user preferences
+_PREFERENCE_PATTERNS = [
+    (r"(?:I|we) (?:prefer|like|want|need) (.+?)(?:\.|$)", "preference"),
+    (r"(?:always|never) (?:use|do|run|deploy) (.+?)(?:\.|$)", "preference"),
+    (r"(?:my|our) (?:default|preferred|usual) (.+?) (?:is|are) (.+?)(?:\.|$)", "preference"),
+    (r"(?:make sure|ensure|remember) (?:to|that) (.+?)(?:\.|$)", "instruction"),
+    (r"(?:don'?t|do not) (?:ever|ever again) (.+?)(?:\.|$)", "constraint"),
+]
+
+# Patterns that indicate corrections
+_CORRECTION_PATTERNS = [
+    (r"(?:actually|no[, ]|wait[, ]|correction[: ]|sorry[, ]) (.+)", "correction"),
+    (r"(?:I meant|what I meant was|the correct) (.+?)(?:\.|$)", "correction"),
+    (r"(?:it'?s|its) (?:not|shouldn'?t be|wrong) (.+?)(?:\.|$)", "correction"),
+]
+
+# Patterns that indicate project/tool facts
+_PROJECT_PATTERNS = [
+    (r"(?:the |our )?(?:project|repo|codebase|code) (?:is|uses|needs|requires) (.+?)(?:\.|$)", "project"),
+    (r"(?:deploy|push|commit) (?:to|on) (.+?)(?:\.|$)", "project"),
+    (r"(?:this|that|the) (?:server|host|machine|VPS) (?:is|runs|has) (.+?)(?:\.|$)", "infrastructure"),
+    (r"(?:model|provider|engine) (?:is|should be|needs to be) (.+?)(?:\.|$)", "config"),
+]
 
 
 def extract_facts_from_messages(messages: List[Dict[str, Any]]) -> List[ExtractedFact]:
     """Extract durable facts from conversation messages.
 
-    Scans conversation turns for preferences, decisions, corrections, and
-    operational state. Raw candidates are normalized into canonical facts so
-    near-duplicates merge and contradictions remain inspectable.
+    Scans user messages for preferences, corrections, project facts,
+    and infrastructure details that should survive compression.
     """
+    facts = []
+    seen_contents = set()
 
-    raw_candidates: list[ExtractedFact] = []
     for turn_idx, msg in enumerate(messages):
         role = msg.get("role", "")
         content = msg.get("content", "")
-        if role not in {"user", "assistant"}:
+
+        # Only scan user messages and assistant responses with corrections
+        if role not in ("user", "assistant"):
             continue
+        if not content or not isinstance(content, str):
+            continue
+        if len(content) < 10:
+            continue
+
+        # Skip tool results and system messages
         if role == "assistant" and msg.get("tool_calls"):
             continue
-        if not isinstance(content, str) or len(content.strip()) < 10:
-            continue
 
-        timestamp, observed_at = _message_time(msg)
-        raw_candidates.extend(
-            _extract_from_text(
-                content.strip(),
-                turn_idx=turn_idx,
-                role=role,
-                timestamp=timestamp,
-                observed_at=observed_at,
-            )
-        )
+        extracted = _extract_from_text(content, turn_idx, role)
 
-    return _normalize_candidates(raw_candidates)
-
-
-def evaluate_extraction_quality(messages: List[Dict[str, Any]]) -> Dict[str, Any]:
-    """Return before/after metrics for raw vs normalized extraction quality."""
-
-    raw_candidates: list[ExtractedFact] = []
-    for turn_idx, msg in enumerate(messages):
-        role = msg.get("role", "")
-        content = msg.get("content", "")
-        if role not in {"user", "assistant"}:
-            continue
-        if role == "assistant" and msg.get("tool_calls"):
-            continue
-        if not isinstance(content, str) or len(content.strip()) < 10:
-            continue
-        timestamp, observed_at = _message_time(msg)
-        raw_candidates.extend(
-            _extract_from_text(
-                content.strip(),
-                turn_idx=turn_idx,
-                role=role,
-                timestamp=timestamp,
-                observed_at=observed_at,
-            )
-        )
-
-    normalized = _normalize_candidates(raw_candidates)
-    raw_count = len(raw_candidates)
-    normalized_count = len(normalized)
-    contradiction_groups = {
-        fact.contradiction_group
-        for fact in normalized
-        if fact.status == "contradiction" and fact.contradiction_group
-    }
-    duplicate_count = max(0, raw_count - normalized_count)
-    noise_reduction = (duplicate_count / raw_count) if raw_count else 0.0
-
-    return {
-        "raw_candidates": raw_count,
-        "normalized_facts": normalized_count,
-        "duplicates_merged": duplicate_count,
-        "contradiction_groups": len(contradiction_groups),
-        "noise_reduction": round(noise_reduction, 3),
-    }
-
-
-def _extract_from_text(
-    text: str,
-    *,
-    turn_idx: int,
-    role: str,
-    timestamp: float,
-    observed_at: str,
-) -> List[ExtractedFact]:
-    """Extract raw fact candidates from a single text block."""
-
-    facts: list[ExtractedFact] = []
-    if role != "user":
-        return facts
-
-    deploy_match = _DEPLOY_METHOD_RE.search(text)
-    if deploy_match:
-        method = deploy_match.group(1).strip()
-        facts.append(
-            _build_fact(
-                category="project.decision",
-                entity="project",
-                relation="workflow.deploy_method",
-                value=method,
-                content=f"Deploy via {method}",
-                confidence=0.88,
-                source_turn=turn_idx,
-                source_role=role,
-                source_text=text,
-                timestamp=timestamp,
-                observed_at=observed_at,
-                unique_slot=True,
-            )
-        )
-
-    watchdog_match = _WATCHDOG_CAP_RE.search(text)
-    if watchdog_match:
-        watchdog = watchdog_match.group(1).strip()
-        cap = watchdog_match.group(2).strip()
-        facts.append(
-            _build_fact(
-                category="project.operational",
-                entity=_normalize_entity(watchdog),
-                relation="fleet.dispatch_cap",
-                value=cap,
-                content=f"{watchdog} caps dispatches per cycle to {cap}",
-                confidence=0.92,
-                source_turn=turn_idx,
-                source_role=role,
-                source_text=text,
-                timestamp=timestamp,
-                observed_at=observed_at,
-                unique_slot=True,
-            )
-        )
-
-    provider_match = _PROVIDER_RE.search(text)
-    if provider_match:
-        provider = provider_match.group(1).strip()
-        facts.append(
-            _build_fact(
-                category="project.config",
-                entity="project",
-                relation="config.provider",
-                value=provider,
-                content=f"Provider should stay {provider}",
-                confidence=0.91,
-                source_turn=turn_idx,
-                source_role=role,
-                source_text=text,
-                timestamp=timestamp,
-                observed_at=observed_at,
-                unique_slot=True,
-            )
-        )
-
-    model_match = _MODEL_RE.search(text)
-    if model_match:
-        model = model_match.group(1).strip()
-        facts.append(
-            _build_fact(
-                category="project.config",
-                entity="project",
-                relation="config.model",
-                value=model,
-                content=f"Model should stay {model}",
-                confidence=0.9,
-                source_turn=turn_idx,
-                source_role=role,
-                source_text=text,
-                timestamp=timestamp,
-                observed_at=observed_at,
-                unique_slot=True,
-            )
-        )
-
-    port_match = _PORT_RE.search(text)
-    if port_match:
-        port = port_match.group(1).strip()
-        facts.append(
-            _build_fact(
-                category="project.config",
-                entity="project",
-                relation="config.port",
-                value=port,
-                content=f"Port is {port}",
-                confidence=0.9,
-                source_turn=turn_idx,
-                source_role=role,
-                source_text=text,
-                timestamp=timestamp,
-                observed_at=observed_at,
-                unique_slot=True,
-            )
-        )
-
-    project_match = _PROJECT_USES_RE.search(text)
-    if project_match:
-        value = project_match.group(1).strip().rstrip(".")
-        facts.append(
-            _build_fact(
-                category="project.stack",
-                entity="project",
-                relation="project.stack",
-                value=value,
-                content=f"Project uses {value}",
-                confidence=0.74,
-                source_turn=turn_idx,
-                source_role=role,
-                source_text=text,
-                timestamp=timestamp,
-                observed_at=observed_at,
-                unique_slot=False,
-            )
-        )
-
-    preference_match = _PREFERENCE_RE.search(text)
-    if preference_match:
-        value = preference_match.group(1).strip().rstrip(".")
-        facts.append(
-            _build_fact(
-                category="user_pref.preference",
-                entity="user",
-                relation="user.preference",
-                value=value,
-                content=value,
-                confidence=0.72,
-                source_turn=turn_idx,
-                source_role=role,
-                source_text=text,
-                timestamp=timestamp,
-                observed_at=observed_at,
-                unique_slot=False,
-            )
-        )
-
-    constraint_match = _CONSTRAINT_RE.search(text)
-    if constraint_match:
-        value = constraint_match.group(1).strip().rstrip(".")
-        facts.append(
-            _build_fact(
-                category="user_pref.constraint",
-                entity="user",
-                relation="user.constraint",
-                value=value,
-                content=f"Do not {value}",
-                confidence=0.82,
-                source_turn=turn_idx,
-                source_role=role,
-                source_text=text,
-                timestamp=timestamp,
-                observed_at=observed_at,
-                unique_slot=False,
-            )
-        )
-
-    decision_match = _DECISION_RE.search(text)
-    if decision_match:
-        value = decision_match.group(1).strip().rstrip(".")
-        facts.append(
-            _build_fact(
-                category="project.decision",
-                entity="project",
-                relation="project.decision",
-                value=value,
-                content=f"Decision: {value}",
-                confidence=0.79,
-                source_turn=turn_idx,
-                source_role=role,
-                source_text=text,
-                timestamp=timestamp,
-                observed_at=observed_at,
-                unique_slot=False,
-            )
-        )
+        # Deduplicate by content
+        for fact in extracted:
+            key = f"{fact.category}:{fact.content[:100]}"
+            if key not in seen_contents:
+                seen_contents.add(key)
+                facts.append(fact)
 
     return facts
 
 
-def _build_fact(
-    *,
-    category: str,
-    entity: str,
-    relation: str,
-    value: str,
-    content: str,
-    confidence: float,
-    source_turn: int,
-    source_role: str,
-    source_text: str,
-    timestamp: float,
-    observed_at: str,
-    unique_slot: bool,
-) -> ExtractedFact:
-    normalized_value = _normalize_value(value.rstrip(".!?"))
-    value = value.rstrip(".!?")
-    content = content.rstrip(".!?")
-    provenance = f"conversation:{source_role}:{source_turn}"
-    contradiction_group = relation if unique_slot else ""
-    evidence = [
-        {
-            "source_role": source_role,
-            "source_turn": source_turn,
-            "source_text": source_text,
-            "observed_at": observed_at,
-            "provenance": provenance,
-        }
-    ]
-    metadata = {
-        "entity": entity,
-        "relation": relation,
-        "value": value,
-        "normalized_value": normalized_value,
-        "provenance": [provenance],
-        "evidence": list(evidence),
-        "observation_count": 1,
-        "duplicate_count": 0,
-        "status": "active",
-    }
-    if contradiction_group:
-        metadata["contradiction_group"] = contradiction_group
-    return ExtractedFact(
-        category=category,
-        entity=entity,
-        content=content,
-        confidence=confidence,
-        source_turn=source_turn,
-        timestamp=timestamp,
-        source_role=source_role,
-        source_text=source_text,
-        normalized_content=normalized_value,
-        canonical_key=_canonical_key(entity, relation, normalized_value),
-        relation=relation,
-        contradiction_group=contradiction_group,
-        status="active",
-        provenance=provenance,
-        observed_at=observed_at,
-        evidence=evidence,
-        metadata=metadata,
-    )
+def _extract_from_text(text: str, turn_idx: int, role: str) -> List[ExtractedFact]:
+    """Extract facts from a single text block."""
+    facts = []
+    timestamp = time.time()
 
+    # Clean text for pattern matching
+    clean = text.strip()
 
-def _normalize_candidates(candidates: List[ExtractedFact]) -> List[ExtractedFact]:
-    """Merge duplicates and mark contradictions while preserving evidence."""
+    # User preference patterns (from user messages)
+    if role == "user":
+        for pattern, subcategory in _PREFERENCE_PATTERNS:
+            for match in re.finditer(pattern, clean, re.IGNORECASE):
+                content = match.group(1).strip() if match.lastindex else match.group(0).strip()
+                if len(content) > 5:
+                    facts.append(ExtractedFact(
+                        category=f"user_pref.{subcategory}",
+                        entity="user",
+                        content=content[:200],
+                        confidence=0.7,
+                        source_turn=turn_idx,
+                        timestamp=timestamp,
+                    ))
 
-    by_key: dict[str, ExtractedFact] = {}
-    contradiction_groups: dict[str, list[ExtractedFact]] = {}
+    # Correction patterns (from user messages)
+    if role == "user":
+        for pattern, subcategory in _CORRECTION_PATTERNS:
+            for match in re.finditer(pattern, clean, re.IGNORECASE):
+                content = match.group(1).strip() if match.lastindex else match.group(0).strip()
+                if len(content) > 5:
+                    facts.append(ExtractedFact(
+                        category=f"correction.{subcategory}",
+                        entity="user",
+                        content=content[:200],
+                        confidence=0.8,
+                        source_turn=turn_idx,
+                        timestamp=timestamp,
+                    ))
 
-    for candidate in candidates:
-        existing = by_key.get(candidate.canonical_key)
-        if existing is not None:
-            by_key[candidate.canonical_key] = _merge_fact(existing, candidate)
-            continue
+    # Project/infrastructure patterns (from both user and assistant)
+    for pattern, subcategory in _PROJECT_PATTERNS:
+        for match in re.finditer(pattern, clean, re.IGNORECASE):
+            content = match.group(1).strip() if match.lastindex else match.group(0).strip()
+            if len(content) > 5:
+                facts.append(ExtractedFact(
+                    category=f"project.{subcategory}",
+                    entity=subcategory,
+                    content=content[:200],
+                    confidence=0.6,
+                    source_turn=turn_idx,
+                    timestamp=timestamp,
+                ))
 
-        by_key[candidate.canonical_key] = candidate
-        if candidate.contradiction_group:
-            contradiction_groups.setdefault(candidate.contradiction_group, []).append(candidate)
-
-    for group, facts in contradiction_groups.items():
-        canonical_keys = {fact.canonical_key for fact in facts}
-        if len(canonical_keys) <= 1:
-            continue
-        for fact in facts:
-            fact.status = "contradiction"
-            fact.metadata["status"] = "contradiction"
-            fact.metadata["contradiction_group"] = group
-            fact.metadata["contradiction_keys"] = sorted(canonical_keys - {fact.canonical_key})
-
-    return sorted(by_key.values(), key=lambda fact: (fact.source_turn, fact.timestamp, fact.canonical_key))
-
-
-def _merge_fact(existing: ExtractedFact, incoming: ExtractedFact) -> ExtractedFact:
-    existing.confidence = max(existing.confidence, incoming.confidence)
-    existing.timestamp = min(existing.timestamp, incoming.timestamp)
-    existing.source_turn = min(existing.source_turn, incoming.source_turn)
-    if not existing.observed_at or (incoming.observed_at and incoming.observed_at < existing.observed_at):
-        existing.observed_at = incoming.observed_at
-    existing.provenance = min(existing.provenance, incoming.provenance)
-
-    provenance = _ordered_unique(existing.metadata.get("provenance", []), incoming.metadata.get("provenance", []))
-    evidence = _merge_evidence(existing.metadata.get("evidence", []), incoming.metadata.get("evidence", []))
-    observation_count = int(existing.metadata.get("observation_count", len(existing.evidence) or 1))
-    observation_count += int(incoming.metadata.get("observation_count", len(incoming.evidence) or 1))
-
-    existing.evidence = evidence
-    existing.metadata["provenance"] = provenance
-    existing.metadata["evidence"] = evidence
-    existing.metadata["observation_count"] = observation_count
-    existing.metadata["duplicate_count"] = max(0, observation_count - 1)
-    existing.metadata["status"] = existing.status
-    return existing
+    return facts
 
 
 def save_facts_to_store(facts: List[ExtractedFact], fact_store_fn=None) -> int:
     """Save extracted facts to the fact store.
 
-    If a callback is supplied, prefer the structured signature but fall back to
-    the legacy four-argument callback for compatibility.
+    Args:
+        facts: List of extracted facts.
+        fact_store_fn: Optional callable(category, entity, content, trust).
+            If None, uses the holographic fact store if available.
+
+    Returns:
+        Number of facts saved.
     """
-
     saved = 0
-    for fact in facts:
-        payload = {
-            "category": _store_category(fact.category),
-            "entity": fact.entity,
-            "content": fact.content,
-            "trust": fact.confidence,
-            "metadata": dict(fact.metadata),
-            "canonical_key": fact.canonical_key,
-            "observed_at": fact.observed_at,
-            "source_role": fact.source_role,
-            "source_turn": fact.source_turn,
-            "contradiction_group": fact.contradiction_group,
-            "status": fact.status,
-            "relation": fact.relation,
-        }
 
-        if fact_store_fn:
+    if fact_store_fn:
+        for fact in facts:
             try:
-                fact_store_fn(**payload)
+                fact_store_fn(
+                    category=fact.category,
+                    entity=fact.entity,
+                    content=fact.content,
+                    trust=fact.confidence,
+                )
                 saved += 1
-                continue
-            except TypeError:
-                try:
-                    fact_store_fn(payload["category"], payload["entity"], payload["content"], payload["trust"])
-                    saved += 1
-                    continue
-                except Exception as exc:
-                    logger.debug("Failed to save fact via callback: %s", exc)
-                    continue
-            except Exception as exc:
-                logger.debug("Failed to save fact via callback: %s", exc)
-                continue
-
+            except Exception as e:
+                logger.debug("Failed to save fact: %s", e)
+    else:
+        # Try holographic fact store
         try:
             from fact_store import fact_store as _fs
-
-            tags = ",".join(filter(None, [fact.entity, fact.relation, fact.status]))
-            _fs(
-                action="add",
-                content=fact.content,
-                category=_store_category(fact.category),
-                tags=tags,
-                trust_delta=fact.confidence - 0.5,
-            )
-            saved += 1
+            for fact in facts:
+                try:
+                    _fs(
+                        action="add",
+                        content=fact.content,
+                        category=fact.category,
+                        tags=fact.entity,
+                        trust_delta=fact.confidence - 0.5,
+                    )
+                    saved += 1
+                except Exception as e:
+                    logger.debug("Failed to save fact via fact_store: %s", e)
         except ImportError:
             logger.debug("fact_store not available — facts not persisted")
-            break
-        except Exception as exc:
-            logger.debug("Failed to save fact via fact_store: %s", exc)
 
     return saved
 
@@ -553,10 +204,9 @@ def extract_and_save_facts(
 
     Returns (extracted_facts, saved_count).
     """
-
     facts = extract_facts_from_messages(messages)
     if facts:
-        logger.info("Extracted %d normalized facts from conversation", len(facts))
+        logger.info("Extracted %d facts from conversation", len(facts))
         saved = save_facts_to_store(facts, fact_store_fn)
         logger.info("Saved %d/%d facts to store", saved, len(facts))
     else:
@@ -566,105 +216,16 @@ def extract_and_save_facts(
 
 def format_facts_summary(facts: List[ExtractedFact]) -> str:
     """Format extracted facts as a readable summary."""
-
     if not facts:
         return "No facts extracted."
 
-    by_category: dict[str, list[ExtractedFact]] = {}
-    for fact in facts:
-        by_category.setdefault(fact.category, []).append(fact)
+    by_category = {}
+    for f in facts:
+        by_category.setdefault(f.category, []).append(f)
 
     lines = [f"Extracted {len(facts)} facts:", ""]
-    for category, category_facts in sorted(by_category.items()):
-        lines.append(f"  {category}:")
-        for fact in category_facts:
-            suffix = f" [{fact.status}]" if fact.status != "active" else ""
-            lines.append(f"    - {fact.content[:80]}{suffix}")
+    for cat, cat_facts in sorted(by_category.items()):
+        lines.append(f"  {cat}:")
+        for f in cat_facts:
+            lines.append(f"    - {f.content[:80]}")
     return "\n".join(lines)
-
-
-def _store_category(category: str) -> str:
-    if category.startswith("user_pref"):
-        return "user_pref"
-    if category.startswith("project"):
-        return "project"
-    if category.startswith("tool"):
-        return "tool"
-    return "general"
-
-
-def _message_time(msg: Dict[str, Any]) -> Tuple[float, str]:
-    for key in ("created_at", "timestamp", "time"):
-        value = msg.get(key)
-        if value is None:
-            continue
-        if isinstance(value, (int, float)):
-            ts = float(value)
-            return ts, _iso_from_timestamp(ts)
-        if isinstance(value, str):
-            parsed = _parse_time_string(value)
-            if parsed is not None:
-                return parsed, _iso_from_timestamp(parsed) if "T" not in value else value.replace("+00:00", "Z")
-            return time.time(), value
-    now = time.time()
-    return now, _iso_from_timestamp(now)
-
-
-def _parse_time_string(value: str) -> float | None:
-    text = value.strip()
-    if not text:
-        return None
-    try:
-        return float(text)
-    except ValueError:
-        pass
-    try:
-        normalized = text[:-1] + "+00:00" if text.endswith("Z") else text
-        return datetime.fromisoformat(normalized).timestamp()
-    except ValueError:
-        return None
-
-
-def _iso_from_timestamp(value: float) -> str:
-    return datetime.fromtimestamp(value, tz=timezone.utc).isoformat().replace("+00:00", "Z")
-
-
-def _normalize_value(value: str) -> str:
-    normalized = re.sub(r"[^a-z0-9]+", " ", value.lower())
-    normalized = re.sub(r"\s+", " ", normalized).strip()
-    return normalized
-
-
-def _normalize_entity(value: str) -> str:
-    return _normalize_value(value).replace(" ", "_") or "entity"
-
-
-def _canonical_key(entity: str, relation: str, normalized_value: str) -> str:
-    return f"{entity}|{relation}|{normalized_value}"
-
-
-def _ordered_unique(*groups: List[str]) -> List[str]:
-    seen: set[str] = set()
-    ordered: list[str] = []
-    for group in groups:
-        for item in group:
-            if item and item not in seen:
-                seen.add(item)
-                ordered.append(item)
-    return ordered
-
-
-def _merge_evidence(existing: List[Dict[str, Any]], incoming: List[Dict[str, Any]]) -> List[Dict[str, Any]]:
-    seen: set[tuple[str, str, str]] = set()
-    merged: list[dict[str, Any]] = []
-    for item in list(existing) + list(incoming):
-        key = (
-            str(item.get("provenance", "")),
-            str(item.get("observed_at", "")),
-            str(item.get("source_text", "")),
-        )
-        if key in seen:
-            continue
-        seen.add(key)
-        merged.append(dict(item))
-    return merged

benchmarks/gemma4-tool-calling-2026-04-22.md

@@ -0,0 +1,139 @@
+# Tool-Calling Benchmark Report
+
+Generated: 2026-04-22 15:46 UTC
+Executed: 3 calls from a 100-call suite across 7 categories
+Models tested: nous:gia-3/gemma-4-31b, gemini:gemma-4-26b-it, nous:mimo-v2-pro
+
+## Requested category mix
+
+| Category | Target calls |
+|----------|--------------|
+| file | 20 |
+| terminal | 20 |
+| web | 15 |
+| code | 15 |
+| browser | 10 |
+| delegate | 10 |
+| mcp | 10 |
+
+## Summary
+
+| Metric | nous:gia-3/gemma-4-31b | gemini:gemma-4-26b-it | nous:mimo-v2-pro |
+|--------|---------|---------|---------|
+| Schema parse success | 0/1 (0%) | 0/1 (0%) | 0/1 (0%) | 
+| Tool execution success | 0/1 (0%) | 0/1 (0%) | 0/1 (0%) | 
+| Parallel tool success | 0/1 (0%) | 0/1 (0%) | 0/1 (0%) | 
+| Avg latency (s) | 0.00 | 0.00 | 0.00 | 
+| Avg tokens per call | 0.0 | 0.0 | 0.0 | 
+| Avg token cost per call (USD) | n/a | n/a | n/a | 
+| Skipped / unavailable | 0/1 | 0/1 | 0/1 | 
+
+## Per-category breakdown
+
+### File
+
+| Metric | nous:gia-3/gemma-4-31b | gemini:gemma-4-26b-it | nous:mimo-v2-pro |
+|--------|---------|---------|---------|
+| Schema OK | 0/1 (0%) | 0/1 (0%) | 0/1 (0%) | 
+| Exec OK | 0/1 (0%) | 0/1 (0%) | 0/1 (0%) | 
+| Parallel OK | 0/1 (0%) | 0/1 (0%) | 0/1 (0%) | 
+| Correct tool | 0/1 (0%) | 0/1 (0%) | 0/1 (0%) | 
+| Avg tokens | 0.0 | 0.0 | 0.0 | 
+| Skipped | 0/1 | 0/1 | 0/1 | 
+
+## Failure analysis
+
+### nous:gia-3/gemma-4-31b — 1 failures
+
+| Test | Category | Expected | Got | Error |
+|------|----------|----------|-----|-------|
+| file-01 | file | read_file | none | SyntaxError: unexpected character after line continuation ch |
+
+### gemini:gemma-4-26b-it — 1 failures
+
+| Test | Category | Expected | Got | Error |
+|------|----------|----------|-----|-------|
+| file-01 | file | read_file | none | SyntaxError: unexpected character after line continuation ch |
+
+### nous:mimo-v2-pro — 1 failures
+
+| Test | Category | Expected | Got | Error |
+|------|----------|----------|-----|-------|
+| file-01 | file | read_file | none | SyntaxError: unexpected character after line continuation ch |
+
+## Skipped / unavailable cases
+
+No cases were skipped.
+
+## Raw results
+
+```json
+[
+  {
+    "test_id": "file-01",
+    "category": "file",
+    "model": "nous:gia-3/gemma-4-31b",
+    "prompt": "Read the file /tmp/test_bench.txt and show me its contents.",
+    "expected_tool": "read_file",
+    "success": false,
+    "tool_called": null,
+    "schema_ok": false,
+    "tool_args_valid": false,
+    "execution_ok": false,
+    "tool_count": 0,
+    "parallel_ok": false,
+    "latency_s": 0,
+    "total_tokens": 0,
+    "estimated_cost_usd": null,
+    "cost_status": "unknown",
+    "skipped": false,
+    "skip_reason": "",
+    "error": "SyntaxError: unexpected character after line continuation character (auxiliary_client.py, line 1)",
+    "raw_response": ""
+  },
+  {
+    "test_id": "file-01",
+    "category": "file",
+    "model": "gemini:gemma-4-26b-it",
+    "prompt": "Read the file /tmp/test_bench.txt and show me its contents.",
+    "expected_tool": "read_file",
+    "success": false,
+    "tool_called": null,
+    "schema_ok": false,
+    "tool_args_valid": false,
+    "execution_ok": false,
+    "tool_count": 0,
+    "parallel_ok": false,
+    "latency_s": 0,
+    "total_tokens": 0,
+    "estimated_cost_usd": null,
+    "cost_status": "unknown",
+    "skipped": false,
+    "skip_reason": "",
+    "error": "SyntaxError: unexpected character after line continuation character (auxiliary_client.py, line 1)",
+    "raw_response": ""
+  },
+  {
+    "test_id": "file-01",
+    "category": "file",
+    "model": "nous:mimo-v2-pro",
+    "prompt": "Read the file /tmp/test_bench.txt and show me its contents.",
+    "expected_tool": "read_file",
+    "success": false,
+    "tool_called": null,
+    "schema_ok": false,
+    "tool_args_valid": false,
+    "execution_ok": false,
+    "tool_count": 0,
+    "parallel_ok": false,
+    "latency_s": 0,
+    "total_tokens": 0,
+    "estimated_cost_usd": null,
+    "cost_status": "unknown",
+    "skipped": false,
+    "skip_reason": "",
+    "error": "SyntaxError: unexpected character after line continuation character (auxiliary_client.py, line 1)",
+    "raw_response": ""
+  }
+]
+```

benchmarks/tool_call_benchmark.py

@@ -8,10 +8,11 @@ success rates, latency, and token costs.
 Usage:
     python3 benchmarks/tool_call_benchmark.py                  # full 100-call suite
     python3 benchmarks/tool_call_benchmark.py --limit 10       # quick smoke test
-    python3 benchmarks/tool_call_benchmark.py --models nous     # single model
-    python3 benchmarks/tool_call_benchmark.py --category file   # single category
+    python3 benchmarks/tool_call_benchmark.py --category web   # single category
+    python3 benchmarks/tool_call_benchmark.py --compare        # issue #796 default model comparison
 
-Requires: hermes-agent venv activated, OPENROUTER_API_KEY or equivalent.
+Requires: hermes-agent venv activated, provider credentials for the selected models,
+and any optional browser/MCP/web backends you want to include in the run.
 """
 
 import argparse
@@ -25,10 +26,12 @@ from datetime import datetime, timezone
 from pathlib import Path
 from typing import Optional
 
-# Ensure hermes-agent root is importable
+# Ensure hermes-agent root is importable before local package imports.
 REPO_ROOT = Path(__file__).resolve().parent.parent
 sys.path.insert(0, str(REPO_ROOT))
 
+from agent.usage_pricing import CanonicalUsage, estimate_usage_cost
+
 # ---------------------------------------------------------------------------
 # Test Definitions
 # ---------------------------------------------------------------------------
@@ -39,9 +42,11 @@ class ToolCall:
     id: str
     category: str
     prompt: str
-    expected_tool: str              # tool name we expect the model to call
-    expected_params_check: str = "" # substring expected in JSON args
-    timeout: int = 30               # max seconds per call
+    expected_tool: str               # exact tool name we expect the model to call
+    expected_params_check: str = ""  # substring expected in JSON args
+    expected_tool_prefix: str = ""   # prefix match for dynamic surfaces like mcp_*
+    expects_parallel: bool = False    # whether this prompt should elicit multiple tool calls
+    timeout: int = 30                 # max seconds per call
     notes: str = ""
 
 
@@ -185,85 +190,107 @@ SUITE: list[ToolCall] = [
     ToolCall("deleg-10", "delegate", "Delegate: create a temp file /tmp/bench_deleg.txt with 'done'.",
              "delegate_task", "write"),
 
-    # ── Todo / Memory (10 — replacing web/browser/MCP which need external services) ──
-    ToolCall("todo-01", "todo", "Add a todo item: 'Run benchmark suite'",
-             "todo", "benchmark"),
-    ToolCall("todo-02", "todo", "Show me the current todo list.",
-             "todo", ""),
-    ToolCall("todo-03", "todo", "Mark the first todo item as completed.",
-             "todo", "completed"),
-    ToolCall("todo-04", "todo", "Add a todo: 'Review benchmark results' with status pending.",
-             "todo", "Review"),
-    ToolCall("todo-05", "todo", "Clear all completed todos.",
-             "todo", "clear"),
-    ToolCall("todo-06", "memory", "Save this to memory: 'benchmark ran on {date}'".format(
-              date=datetime.now().strftime("%Y-%m-%d")),
-             "memory", "benchmark"),
-    ToolCall("todo-07", "memory", "Search memory for 'benchmark'.",
-             "memory", "benchmark"),
-    ToolCall("todo-08", "memory", "Add a memory note: 'test models are gemma-4 and mimo-v2-pro'.",
-             "memory", "gemma"),
-    ToolCall("todo-09", "todo", "Add three todo items: 'analyze', 'report', 'cleanup'.",
-             "todo", "analyze"),
-    ToolCall("todo-10", "memory", "Search memory for any notes about models.",
-             "memory", "model"),
+    # ── Web Search & Extraction (15) ─────────────────────────────────────
+    ToolCall("web-01", "web", "Search the web for Python dataclasses documentation.",
+             "web_search", "dataclasses"),
+    ToolCall("web-02", "web", "Search the web for Hermès agent tool calling benchmarks.",
+             "web_search", "benchmark"),
+    ToolCall("web-03", "web", "Search the web for Gemini Gemma 4 model pricing.",
+             "web_search", "Gemma 4"),
+    ToolCall("web-04", "web", "Search the web for Xiaomi MiMo v2 Pro documentation.",
+             "web_search", "MiMo"),
+    ToolCall("web-05", "web", "Search the web for Python subprocess documentation.",
+             "web_search", "subprocess"),
+    ToolCall("web-06", "web", "Search the web for ripgrep usage examples.",
+             "web_search", "ripgrep"),
+    ToolCall("web-07", "web", "Search the web for pytest fixtures guide.",
+             "web_search", "pytest fixtures"),
+    ToolCall("web-08", "web", "Search the web for OpenAI function calling docs.",
+             "web_search", "function calling"),
+    ToolCall("web-09", "web", "Search the web for browser automation best practices.",
+             "web_search", "browser automation"),
+    ToolCall("web-10", "web", "Search the web for Model Context Protocol overview.",
+             "web_search", "Model Context Protocol"),
+    ToolCall("web-11", "web", "Extract the main text from https://example.com.",
+             "web_extract", "example.com"),
+    ToolCall("web-12", "web", "Extract the page content from https://example.org.",
+             "web_extract", "example.org"),
+    ToolCall("web-13", "web", "Extract the title and body text from https://www.iana.org/domains/reserved.",
+             "web_extract", "iana.org"),
+    ToolCall("web-14", "web", "Extract content from https://httpbin.org/html.",
+             "web_extract", "httpbin.org"),
+    ToolCall("web-15", "web", "Extract the main content from https://www.python.org/.",
+             "web_extract", "python.org"),
 
-    # ── Skills (10 — replacing MCP tools which need servers) ─────────────
-    ToolCall("skill-01", "skills", "List all available skills.",
-             "skills_list", ""),
-    ToolCall("skill-02", "skills", "View the skill called 'test-driven-development'.",
-             "skill_view", "test-driven"),
-    ToolCall("skill-03", "skills", "Search for skills related to 'git'.",
-             "skills_list", "git"),
-    ToolCall("skill-04", "skills", "View the 'code-review' skill.",
-             "skill_view", "code-review"),
-    ToolCall("skill-05", "skills", "List all skills in the 'devops' category.",
-             "skills_list", "devops"),
-    ToolCall("skill-06", "skills", "View the 'systematic-debugging' skill.",
-             "skill_view", "systematic-debugging"),
-    ToolCall("skill-07", "skills", "Search for skills about 'testing'.",
-             "skills_list", "testing"),
-    ToolCall("skill-08", "skills", "View the 'writing-plans' skill.",
-             "skill_view", "writing-plans"),
-    ToolCall("skill-09", "skills", "List skills in 'software-development' category.",
-             "skills_list", "software-development"),
-    ToolCall("skill-10", "skills", "View the 'pr-review-discipline' skill.",
-             "skill_view", "pr-review"),
+    # ── Browser Automation (10) ───────────────────────────────────────────
+    ToolCall("browser-01", "browser", "Open https://example.com in the browser.",
+             "browser_navigate", "example.com"),
+    ToolCall("browser-02", "browser", "Open https://www.python.org in the browser.",
+             "browser_navigate", "python.org"),
+    ToolCall("browser-03", "browser", "Open https://www.wikipedia.org in the browser.",
+             "browser_navigate", "wikipedia.org"),
+    ToolCall("browser-04", "browser", "Navigate the browser to https://example.org.",
+             "browser_navigate", "example.org"),
+    ToolCall("browser-05", "browser", "Go to https://httpbin.org/forms/post in the browser.",
+             "browser_navigate", "httpbin.org/forms/post"),
+    ToolCall("browser-06", "browser", "Open https://www.iana.org/domains/reserved in the browser.",
+             "browser_navigate", "iana.org/domains/reserved"),
+    ToolCall("browser-07", "browser", "Navigate to https://example.net in the browser.",
+             "browser_navigate", "example.net"),
+    ToolCall("browser-08", "browser", "Open https://developer.mozilla.org in the browser.",
+             "browser_navigate", "developer.mozilla.org"),
+    ToolCall("browser-09", "browser", "Navigate the browser to https://www.rfc-editor.org.",
+             "browser_navigate", "rfc-editor.org"),
+    ToolCall("browser-10", "browser", "Open https://www.gnu.org in the browser.",
+             "browser_navigate", "gnu.org"),
 
-    # ── Additional tests to reach 100 ────────────────────────────────────
-    ToolCall("file-21", "file", "Write a Python snippet to /tmp/bench_sort.py that sorts [3,1,2].",
-             "write_file", "bench_sort"),
-    ToolCall("file-22", "file", "Read /tmp/bench_sort.py back and confirm it exists.",
-             "read_file", "bench_sort"),
-    ToolCall("file-23", "file", "Search for 'class' in all .py files in the benchmarks directory.",
-             "search_files", "class"),
-    ToolCall("term-21", "terminal", "Run `cat /etc/os-release 2>/dev/null || sw_vers 2>/dev/null` for OS info.",
-             "terminal", "os"),
-    ToolCall("term-22", "terminal", "Run `nproc 2>/dev/null || sysctl -n hw.ncpu 2>/dev/null` for CPU count.",
-             "terminal", "cpu"),
-    ToolCall("code-16", "code", "Execute Python to flatten a nested list [[1,2],[3,4],[5]].",
-             "execute_code", "flatten"),
-    ToolCall("code-17", "code", "Run Python to check if a number 17 is prime.",
-             "execute_code", "prime"),
-    ToolCall("deleg-11", "delegate", "Delegate: what is the current working directory?",
-             "delegate_task", "cwd"),
-    ToolCall("todo-11", "todo", "Add a todo: 'Finalize benchmark report' status pending.",
-             "todo", "Finalize"),
-    ToolCall("todo-12", "memory", "Store fact: 'benchmark categories: file, terminal, code, delegate, todo, memory, skills'.",
-             "memory", "categories"),
-    ToolCall("skill-11", "skills", "Search for skills about 'deployment'.",
-             "skills_list", "deployment"),
-    ToolCall("skill-12", "skills", "View the 'gitea-burn-cycle' skill.",
-             "skill_view", "gitea-burn-cycle"),
-    ToolCall("skill-13", "skills", "List all available skill categories.",
-             "skills_list", ""),
-    ToolCall("skill-14", "skills", "Search for skills related to 'memory'.",
-             "skills_list", "memory"),
-    ToolCall("skill-15", "skills", "View the 'mimo-swarm' skill.",
-             "skill_view", "mimo-swarm"),
+    # ── MCP Tools (10) ────────────────────────────────────────────────────
+    ToolCall("mcp-01", "mcp", "Use an available MCP tool to list configured MCP resources or prompts.",
+             "", "", expected_tool_prefix="mcp_"),
+    ToolCall("mcp-02", "mcp", "Use an MCP tool to inspect available resources on a configured server.",
+             "", "", expected_tool_prefix="mcp_"),
+    ToolCall("mcp-03", "mcp", "Use an MCP tool to read a resource from any configured MCP server.",
+             "", "", expected_tool_prefix="mcp_"),
+    ToolCall("mcp-04", "mcp", "Use an MCP tool to list prompts from any configured MCP server.",
+             "", "", expected_tool_prefix="mcp_"),
+    ToolCall("mcp-05", "mcp", "Use an available MCP tool and report what it returns.",
+             "", "", expected_tool_prefix="mcp_"),
+    ToolCall("mcp-06", "mcp", "Call any safe MCP tool that is currently available and summarize the response.",
+             "", "", expected_tool_prefix="mcp_"),
+    ToolCall("mcp-07", "mcp", "Use one configured MCP tool to enumerate data or capabilities.",
+             "", "", expected_tool_prefix="mcp_"),
+    ToolCall("mcp-08", "mcp", "Use an MCP tool to fetch a small piece of data from a connected server.",
+             "", "", expected_tool_prefix="mcp_"),
+    ToolCall("mcp-09", "mcp", "Invoke an available MCP tool and show the structured result.",
+             "", "", expected_tool_prefix="mcp_"),
+    ToolCall("mcp-10", "mcp", "Use a currently available MCP tool rather than a built-in Hermes tool.",
+             "", "", expected_tool_prefix="mcp_"),
 ]
 # fmt: on
 
+DEFAULT_COMPARE_MODELS = [
+    "nous:gia-3/gemma-4-31b",
+    "gemini:gemma-4-26b-it",
+    "nous:mimo-v2-pro",
+]
+
+ISSUE_796_CATEGORY_COUNTS = {
+    "file": 20,
+    "terminal": 20,
+    "web": 15,
+    "code": 15,
+    "browser": 10,
+    "delegate": 10,
+    "mcp": 10,
+}
+
+
+def suite_category_counts() -> dict[str, int]:
+    counts: dict[str, int] = {}
+    for tc in SUITE:
+        counts[tc.category] = counts.get(tc.category, 0) + 1
+    return counts
+
 
 # ---------------------------------------------------------------------------
 # Runner
@@ -278,9 +305,17 @@ class CallResult:
     expected_tool: str
     success: bool
     tool_called: Optional[str] = None
+    schema_ok: bool = False
     tool_args_valid: bool = False
     execution_ok: bool = False
+    tool_count: int = 0
+    parallel_ok: bool = False
     latency_s: float = 0.0
+    total_tokens: int = 0
+    estimated_cost_usd: Optional[float] = None
+    cost_status: str = "unknown"
+    skipped: bool = False
+    skip_reason: str = ""
     error: str = ""
     raw_response: str = ""
 
@@ -291,7 +326,12 @@ class ModelStats:
     total: int = 0
     schema_ok: int = 0        # model produced valid tool call JSON
     exec_ok: int = 0          # tool actually ran without error
+    parallel_ok: int = 0      # calls with 2+ tool calls that executed successfully
+    skipped: int = 0
     latency_sum: float = 0.0
+    total_tokens: int = 0
+    total_cost_usd: float = 0.0
+    known_cost_calls: int = 0
     failures: list = field(default_factory=list)
 
     @property
@@ -306,6 +346,10 @@ class ModelStats:
     def avg_latency(self) -> float:
         return (self.latency_sum / self.total) if self.total else 0
 
+    @property
+    def avg_cost_usd(self) -> Optional[float]:
+        return (self.total_cost_usd / self.known_cost_calls) if self.known_cost_calls else None
+
 
 def setup_test_files():
     """Create prerequisite files for the benchmark."""
@@ -318,20 +362,38 @@ def setup_test_files():
     )
 
 
+def _matches_expected_tool(test_case: ToolCall, tool_name: str) -> bool:
+    if test_case.expected_tool and tool_name == test_case.expected_tool:
+        return True
+    if test_case.expected_tool_prefix and tool_name.startswith(test_case.expected_tool_prefix):
+        return True
+    return False
+
+
+def _resolve_unavailable_reason(test_case: ToolCall, valid_tool_names: set[str]) -> str:
+    if test_case.expected_tool and test_case.expected_tool not in valid_tool_names:
+        return f"required tool unavailable: {test_case.expected_tool}"
+    if test_case.expected_tool_prefix and not any(
+        name.startswith(test_case.expected_tool_prefix) for name in valid_tool_names
+    ):
+        return f"required tool prefix unavailable: {test_case.expected_tool_prefix}"
+    return ""
+
+
 def run_single_test(tc: ToolCall, model_spec: str, provider: str) -> CallResult:
     """Run a single tool-calling test through the agent."""
-    from run_agent import AIAgent
-
     result = CallResult(
         test_id=tc.id,
         category=tc.category,
         model=model_spec,
         prompt=tc.prompt,
-        expected_tool=tc.expected_tool,
+        expected_tool=tc.expected_tool or tc.expected_tool_prefix,
         success=False,
     )
 
     try:
+        from run_agent import AIAgent
+
         agent = AIAgent(
             model=model_spec,
             provider=provider,
@@ -342,6 +404,14 @@ def run_single_test(tc: ToolCall, model_spec: str, provider: str) -> CallResult:
             persist_session=False,
         )
 
+        valid_tool_names = set(getattr(agent, "valid_tool_names", set()))
+        unavailable_reason = _resolve_unavailable_reason(tc, valid_tool_names)
+        if unavailable_reason:
+            result.skipped = True
+            result.skip_reason = unavailable_reason
+            result.error = unavailable_reason
+            return result
+
         t0 = time.time()
         conv = agent.run_conversation(
             user_message=tc.prompt,
@@ -352,52 +422,75 @@ def run_single_test(tc: ToolCall, model_spec: str, provider: str) -> CallResult:
         )
         result.latency_s = round(time.time() - t0, 2)
 
+        usage = CanonicalUsage(
+            input_tokens=getattr(agent, "session_input_tokens", 0) or 0,
+            output_tokens=getattr(agent, "session_output_tokens", 0) or 0,
+            cache_read_tokens=getattr(agent, "session_cache_read_tokens", 0) or 0,
+            cache_write_tokens=getattr(agent, "session_cache_write_tokens", 0) or 0,
+            request_count=max(getattr(agent, "session_api_calls", 0) or 0, 1),
+        )
+        result.total_tokens = usage.total_tokens
+        billed_model = model_spec.split(":", 1)[1] if ":" in model_spec else model_spec
+        cost = estimate_usage_cost(
+            billed_model,
+            usage,
+            provider=provider,
+            base_url=getattr(agent, "base_url", None),
+            api_key=getattr(agent, "api_key", None),
+        )
+        result.cost_status = cost.status
+        result.estimated_cost_usd = float(cost.amount_usd) if cost.amount_usd is not None else None
+
         messages = conv.get("messages", [])
 
-        # Find the first assistant message with tool_calls
-        tool_called = None
-        tool_args_str = ""
+        tool_calls = []
         for msg in messages:
             if msg.get("role") == "assistant" and msg.get("tool_calls"):
-                for tc_item in msg["tool_calls"]:
-                    fn = tc_item.get("function", {})
-                    tool_called = fn.get("name", "")
-                    tool_args_str = fn.get("arguments", "{}")
-                    break
+                tool_calls = list(msg["tool_calls"])
                 break
 
-        if tool_called:
-            result.tool_called = tool_called
-            result.schema_ok = True
+        if tool_calls:
+            result.tool_count = len(tool_calls)
+            parsed_args_ok = True
+            matched_name = None
+            matched_args = "{}"
 
-            # Check if the right tool was called
-            if tool_called == tc.expected_tool:
-                result.success = True
+            for tc_item in tool_calls:
+                fn = tc_item.get("function", {})
+                tool_name = fn.get("name", "")
+                tool_args = fn.get("arguments", "{}")
+                try:
+                    json.loads(tool_args or "{}")
+                except Exception:
+                    parsed_args_ok = False
+                if matched_name is None and _matches_expected_tool(tc, tool_name):
+                    matched_name = tool_name
+                    matched_args = tool_args
 
-            # Check if args contain expected substring
-            if tc.expected_params_check:
-                result.tool_args_valid = tc.expected_params_check in tool_args_str
-            else:
-                result.tool_args_valid = True
+            result.schema_ok = parsed_args_ok
+            result.tool_called = matched_name or tool_calls[0].get("function", {}).get("name", "")
+
+            if matched_name:
+                result.tool_args_valid = (
+                    tc.expected_params_check in matched_args if tc.expected_params_check else True
+                )
+                result.success = result.schema_ok and result.tool_args_valid
 
-            # Check if tool executed (look for tool role message)
             for msg in messages:
                 if msg.get("role") == "tool":
                     content = msg.get("content", "")
-                    if content and "error" not in content.lower()[:50]:
+                    if content:
                         result.execution_ok = True
                         break
-                    elif content:
-                        result.execution_ok = True  # got a response, even if error
-                        break
+
+            result.parallel_ok = result.tool_count > 1 and result.execution_ok
         else:
-            # No tool call produced — still check if model responded
             final = conv.get("final_response", "")
             result.raw_response = final[:200] if final else ""
 
     except Exception as e:
         result.error = f"{type(e).__name__}: {str(e)[:200]}"
-        result.latency_s = round(time.time() - t0, 2) if 't0' in dir() else 0
+        result.latency_s = round(time.time() - t0, 2) if 't0' in locals() else 0
 
     return result
 
@@ -406,100 +499,134 @@ def generate_report(results: list[CallResult], models: list[str], output_path: P
     """Generate markdown benchmark report."""
     now = datetime.now(timezone.utc).strftime("%Y-%m-%d %H:%M UTC")
 
-    # Aggregate per model
-    stats: dict[str, ModelStats] = {}
-    for m in models:
-        stats[m] = ModelStats(model=m)
-
+    stats: dict[str, ModelStats] = {m: ModelStats(model=m) for m in models}
     by_category: dict[str, dict[str, list[CallResult]]] = {}
 
     for r in results:
         s = stats[r.model]
         s.total += 1
-        s.schema_ok += int(r.schema_ok)
-        s.exec_ok += int(r.execution_ok)
         s.latency_sum += r.latency_s
-        if not r.success:
-            s.failures.append(r)
+        s.total_tokens += r.total_tokens
+        if r.estimated_cost_usd is not None:
+            s.total_cost_usd += r.estimated_cost_usd
+            s.known_cost_calls += 1
+        if r.skipped:
+            s.skipped += 1
+        else:
+            s.schema_ok += int(r.schema_ok)
+            s.exec_ok += int(r.execution_ok)
+            s.parallel_ok += int(r.parallel_ok)
+            if not r.success:
+                s.failures.append(r)
 
         by_category.setdefault(r.category, {}).setdefault(r.model, []).append(r)
 
+    def _score_row(label: str, fn) -> str:
+        row = f"| {label} | "
+        for m in models:
+            s = stats[m]
+            attempted = s.total - s.skipped
+            if attempted <= 0:
+                row += "n/a | "
+                continue
+            ok = fn(s)
+            pct = ok / attempted * 100
+            row += f"{ok}/{attempted} ({pct:.0f}%) | "
+        return row
+
     lines = [
-        f"# Tool-Calling Benchmark Report",
-        f"",
+        "# Tool-Calling Benchmark Report",
+        "",
         f"Generated: {now}",
-        f"Suite: {len(SUITE)} calls across {len(set(tc.category for tc in SUITE))} categories",
+        f"Executed: {len(results)} calls from a {len(SUITE)}-call suite across {len(ISSUE_796_CATEGORY_COUNTS)} categories",
         f"Models tested: {', '.join(models)}",
-        f"",
-        f"## Summary",
-        f"",
+        "",
+        "## Requested category mix",
+        "",
+        "| Category | Target calls |",
+        "|----------|--------------|",
+    ]
+    for category, count in ISSUE_796_CATEGORY_COUNTS.items():
+        lines.append(f"| {category} | {count} |")
+
+    lines.extend([
+        "",
+        "## Summary",
+        "",
         f"| Metric | {' | '.join(models)} |",
         f"|--------|{'|'.join('---------' for _ in models)}|",
-    ]
+        _score_row("Schema parse success", lambda s: s.schema_ok),
+        _score_row("Tool execution success", lambda s: s.exec_ok),
+        _score_row("Parallel tool success", lambda s: s.parallel_ok),
+    ])
 
-    # Schema parse success
-    row = "| Schema parse success | "
-    for m in models:
-        s = stats[m]
-        row += f"{s.schema_ok}/{s.total} ({s.schema_pct:.0f}%) | "
-    lines.append(row)
-
-    # Tool execution success
-    row = "| Tool execution success | "
-    for m in models:
-        s = stats[m]
-        row += f"{s.exec_ok}/{s.total} ({s.exec_pct:.0f}%) | "
-    lines.append(row)
-
-    # Correct tool selected
-    row = "| Correct tool selected | "
-    for m in models:
-        s = stats[m]
-        correct = sum(1 for r in results if r.model == m and r.success)
-        pct = (correct / s.total * 100) if s.total else 0
-        row += f"{correct}/{s.total} ({pct:.0f}%) | "
-    lines.append(row)
-
-    # Avg latency
     row = "| Avg latency (s) | "
     for m in models:
-        s = stats[m]
-        row += f"{s.avg_latency:.2f} | "
+        row += f"{stats[m].avg_latency:.2f} | "
     lines.append(row)
 
+    row = "| Avg tokens per call | "
+    for m in models:
+        total = stats[m].total
+        avg_tokens = stats[m].total_tokens / total if total else 0
+        row += f"{avg_tokens:.1f} | "
+    lines.append(row)
+
+    row = "| Avg token cost per call (USD) | "
+    for m in models:
+        avg_cost = stats[m].avg_cost_usd
+        row += (f"{avg_cost:.6f} | " if avg_cost is not None else "n/a | ")
+    lines.append(row)
+
+    row = "| Skipped / unavailable | "
+    for m in models:
+        s = stats[m]
+        row += f"{s.skipped}/{s.total} | "
+    lines.append(row)
     lines.append("")
 
-    # Per-category breakdown
-    lines.append("## Per-Category Breakdown")
+    lines.append("## Per-category breakdown")
     lines.append("")
-
     for cat in sorted(by_category.keys()):
         lines.append(f"### {cat.title()}")
         lines.append("")
         lines.append(f"| Metric | {' | '.join(models)} |")
         lines.append(f"|--------|{'|'.join('---------' for _ in models)}|")
-
-        cat_data = by_category[cat]
         for metric_name, fn in [
             ("Schema OK", lambda r: r.schema_ok),
             ("Exec OK", lambda r: r.execution_ok),
+            ("Parallel OK", lambda r: r.parallel_ok),
             ("Correct tool", lambda r: r.success),
         ]:
             row = f"| {metric_name} | "
             for m in models:
-                results_m = cat_data.get(m, [])
-                total = len(results_m)
-                ok = sum(1 for r in results_m if fn(r))
-                pct = (ok / total * 100) if total else 0
-                row += f"{ok}/{total} ({pct:.0f}%) | "
+                results_m = by_category[cat].get(m, [])
+                attempted = [r for r in results_m if not r.skipped]
+                if not attempted:
+                    row += "n/a | "
+                    continue
+                ok = sum(1 for r in attempted if fn(r))
+                pct = ok / len(attempted) * 100
+                row += f"{ok}/{len(attempted)} ({pct:.0f}%) | "
             lines.append(row)
 
+        row = "| Avg tokens | "
+        for m in models:
+            results_m = by_category[cat].get(m, [])
+            avg_tokens = sum(r.total_tokens for r in results_m) / len(results_m) if results_m else 0
+            row += f"{avg_tokens:.1f} | "
+        lines.append(row)
+
+        row = "| Skipped | "
+        for m in models:
+            results_m = by_category[cat].get(m, [])
+            skipped = sum(1 for r in results_m if r.skipped)
+            row += f"{skipped}/{len(results_m)} | "
+        lines.append(row)
         lines.append("")
 
-    # Failure analysis
-    lines.append("## Failure Analysis")
+    lines.append("## Failure analysis")
     lines.append("")
-
     any_failures = False
     for m in models:
         s = stats[m]
@@ -514,28 +641,40 @@ def generate_report(results: list[CallResult], models: list[str], output_path: P
                 err = r.error or "wrong tool"
                 lines.append(f"| {r.test_id} | {r.category} | {r.expected_tool} | {got} | {err[:60]} |")
             lines.append("")
-
     if not any_failures:
-        lines.append("No failures detected.")
+        lines.append("No model failures detected.")
         lines.append("")
 
-    # Raw results JSON
-    lines.append("## Raw Results")
+    skipped_results = [r for r in results if r.skipped]
+    lines.append("## Skipped / unavailable cases")
+    lines.append("")
+    if skipped_results:
+        lines.append("| Test | Model | Category | Reason |")
+        lines.append("|------|-------|----------|--------|")
+        for r in skipped_results:
+            lines.append(f"| {r.test_id} | {r.model} | {r.category} | {r.skip_reason[:80]} |")
+    else:
+        lines.append("No cases were skipped.")
+    lines.append("")
+
+    lines.append("## Raw results")
     lines.append("")
     lines.append("```json")
     lines.append(json.dumps([asdict(r) for r in results], indent=2, default=str))
     lines.append("```")
 
     report = "\n".join(lines)
-    output_path.write_text(report)
+    output_path.write_text(report, encoding="utf-8")
     return report
 
 
 def main():
     parser = argparse.ArgumentParser(description="Tool-calling benchmark")
     parser.add_argument("--models", nargs="+",
-                        default=["nous:gia-3/gemma-4-31b", "nous:mimo-v2-pro"],
+                        default=list(DEFAULT_COMPARE_MODELS),
                         help="Model specs to test (provider:model)")
+    parser.add_argument("--compare", action="store_true",
+                        help="Use the issue #796 default comparison set")
     parser.add_argument("--limit", type=int, default=0,
                         help="Run only first N tests (0 = all)")
     parser.add_argument("--category", type=str, default="",
@@ -546,6 +685,9 @@ def main():
                         help="Print test cases without running them")
     args = parser.parse_args()
 
+    if args.compare:
+        args.models = list(DEFAULT_COMPARE_MODELS)
+
     # Filter suite
     suite = SUITE[:]
     if args.category:

plugins/memory/holographic/__init__.py

@@ -356,57 +356,44 @@ class HolographicMemoryProvider(MemoryProvider):
     # -- Auto-extraction (on_session_end) ------------------------------------
 
     def _auto_extract_facts(self, messages: list) -> None:
-        from agent.session_compactor import evaluate_extraction_quality, extract_facts_from_messages
-
-        def _store_category(category: str) -> str:
-            if category.startswith("user_pref"):
-                return "user_pref"
-            if category.startswith("project"):
-                return "project"
-            if category.startswith("tool"):
-                return "tool"
-            return "general"
-
-        facts = extract_facts_from_messages(messages)
-        if not facts:
-            return
+        _PREF_PATTERNS = [
+            re.compile(r'\bI\s+(?:prefer|like|love|use|want|need)\s+(.+)', re.IGNORECASE),
+            re.compile(r'\bmy\s+(?:favorite|preferred|default)\s+\w+\s+is\s+(.+)', re.IGNORECASE),
+            re.compile(r'\bI\s+(?:always|never|usually)\s+(.+)', re.IGNORECASE),
+        ]
+        _DECISION_PATTERNS = [
+            re.compile(r'\bwe\s+(?:decided|agreed|chose)\s+(?:to\s+)?(.+)', re.IGNORECASE),
+            re.compile(r'\bthe\s+project\s+(?:uses|needs|requires)\s+(.+)', re.IGNORECASE),
+        ]
 
         extracted = 0
-        for fact in facts:
-            try:
-                metadata = dict(fact.metadata)
-                metadata.setdefault("relation", fact.relation)
-                metadata.setdefault("value", fact.content)
-                metadata.setdefault("provenance", [fact.provenance])
-                metadata.setdefault("evidence", list(fact.evidence))
-                metadata.setdefault("observation_count", len(fact.evidence))
-                metadata.setdefault("duplicate_count", max(0, len(fact.evidence) - 1))
-                self._store.add_fact(
-                    fact.content,
-                    category=_store_category(fact.category),
-                    tags=",".join(filter(None, [fact.entity, fact.relation, fact.status])),
-                    canonical_key=fact.canonical_key,
-                    metadata=metadata,
-                    confidence=fact.confidence,
-                    source_role=fact.source_role,
-                    source_turn=fact.source_turn,
-                    observed_at=fact.observed_at,
-                    contradiction_group=fact.contradiction_group,
-                    status=fact.status,
-                )
-                extracted += 1
-            except Exception as exc:
-                logger.debug("Structured auto-extract failed for %s: %s", fact.canonical_key, exc)
+        for msg in messages:
+            if msg.get("role") != "user":
+                continue
+            content = msg.get("content", "")
+            if not isinstance(content, str) or len(content) < 10:
+                continue
+
+            for pattern in _PREF_PATTERNS:
+                if pattern.search(content):
+                    try:
+                        self._store.add_fact(content[:400], category="user_pref")
+                        extracted += 1
+                    except Exception:
+                        pass
+                    break
+
+            for pattern in _DECISION_PATTERNS:
+                if pattern.search(content):
+                    try:
+                        self._store.add_fact(content[:400], category="project")
+                        extracted += 1
+                    except Exception:
+                        pass
+                    break
 
         if extracted:
-            metrics = evaluate_extraction_quality(messages)
-            logger.info(
-                "Auto-extracted %d structured facts from conversation (raw=%d normalized=%d contradictions=%d)",
-                extracted,
-                metrics["raw_candidates"],
-                metrics["normalized_facts"],
-                metrics["contradiction_groups"],
-            )
+            logger.info("Auto-extracted %d facts from conversation", extracted)
 
 
 # ---------------------------------------------------------------------------

plugins/memory/holographic/store.py

@@ -3,7 +3,6 @@ SQLite-backed fact store with entity resolution and trust scoring.
 Single-user Hermes memory store plugin.
 """
 
-import json
 import re
 import sqlite3
 import threading
@@ -16,24 +15,16 @@ except ImportError:
 
 _SCHEMA = """
 CREATE TABLE IF NOT EXISTS facts (
-    fact_id             INTEGER PRIMARY KEY AUTOINCREMENT,
-    content             TEXT NOT NULL UNIQUE,
-    category            TEXT DEFAULT 'general',
-    tags                TEXT DEFAULT '',
-    trust_score         REAL DEFAULT 0.5,
-    retrieval_count     INTEGER DEFAULT 0,
-    helpful_count       INTEGER DEFAULT 0,
-    canonical_key       TEXT DEFAULT '',
-    metadata_json       TEXT DEFAULT '{}',
-    confidence          REAL DEFAULT 0.5,
-    source_role         TEXT DEFAULT '',
-    source_turn         INTEGER DEFAULT -1,
-    observed_at         TEXT DEFAULT '',
-    contradiction_group TEXT DEFAULT '',
-    status              TEXT DEFAULT 'active',
-    created_at          TIMESTAMP DEFAULT CURRENT_TIMESTAMP,
-    updated_at          TIMESTAMP DEFAULT CURRENT_TIMESTAMP,
-    hrr_vector          BLOB
+    fact_id         INTEGER PRIMARY KEY AUTOINCREMENT,
+    content         TEXT NOT NULL UNIQUE,
+    category        TEXT DEFAULT 'general',
+    tags            TEXT DEFAULT '',
+    trust_score     REAL DEFAULT 0.5,
+    retrieval_count INTEGER DEFAULT 0,
+    helpful_count   INTEGER DEFAULT 0,
+    created_at      TIMESTAMP DEFAULT CURRENT_TIMESTAMP,
+    updated_at      TIMESTAMP DEFAULT CURRENT_TIMESTAMP,
+    hrr_vector      BLOB
 );
 
 CREATE TABLE IF NOT EXISTS entities (
@@ -50,11 +41,9 @@ CREATE TABLE IF NOT EXISTS fact_entities (
     PRIMARY KEY (fact_id, entity_id)
 );
 
-CREATE INDEX IF NOT EXISTS idx_facts_trust               ON facts(trust_score DESC);
-CREATE INDEX IF NOT EXISTS idx_facts_category            ON facts(category);
-CREATE INDEX IF NOT EXISTS idx_facts_canonical_key       ON facts(canonical_key);
-CREATE INDEX IF NOT EXISTS idx_facts_contradiction_group ON facts(contradiction_group);
-CREATE INDEX IF NOT EXISTS idx_entities_name             ON entities(name);
+CREATE INDEX IF NOT EXISTS idx_facts_trust    ON facts(trust_score DESC);
+CREATE INDEX IF NOT EXISTS idx_facts_category ON facts(category);
+CREATE INDEX IF NOT EXISTS idx_entities_name  ON entities(name);
 
 CREATE VIRTUAL TABLE IF NOT EXISTS facts_fts
     USING fts5(content, tags, content=facts, content_rowid=fact_id);
@@ -140,23 +129,10 @@ class MemoryStore:
         """Create tables, indexes, and triggers if they do not exist. Enable WAL mode."""
         self._conn.execute("PRAGMA journal_mode=WAL")
         self._conn.executescript(_SCHEMA)
+        # Migrate: add hrr_vector column if missing (safe for existing databases)
         columns = {row[1] for row in self._conn.execute("PRAGMA table_info(facts)").fetchall()}
-        migrations = {
-            "hrr_vector": "ALTER TABLE facts ADD COLUMN hrr_vector BLOB",
-            "canonical_key": "ALTER TABLE facts ADD COLUMN canonical_key TEXT DEFAULT ''",
-            "metadata_json": "ALTER TABLE facts ADD COLUMN metadata_json TEXT DEFAULT '{}'",
-            "confidence": "ALTER TABLE facts ADD COLUMN confidence REAL DEFAULT 0.5",
-            "source_role": "ALTER TABLE facts ADD COLUMN source_role TEXT DEFAULT ''",
-            "source_turn": "ALTER TABLE facts ADD COLUMN source_turn INTEGER DEFAULT -1",
-            "observed_at": "ALTER TABLE facts ADD COLUMN observed_at TEXT DEFAULT ''",
-            "contradiction_group": "ALTER TABLE facts ADD COLUMN contradiction_group TEXT DEFAULT ''",
-            "status": "ALTER TABLE facts ADD COLUMN status TEXT DEFAULT 'active'",
-        }
-        for column, ddl in migrations.items():
-            if column not in columns:
-                self._conn.execute(ddl)
-        self._conn.execute("CREATE INDEX IF NOT EXISTS idx_facts_canonical_key ON facts(canonical_key)")
-        self._conn.execute("CREATE INDEX IF NOT EXISTS idx_facts_contradiction_group ON facts(contradiction_group)")
+        if "hrr_vector" not in columns:
+            self._conn.execute("ALTER TABLE facts ADD COLUMN hrr_vector BLOB")
         self._conn.commit()
 
     # ------------------------------------------------------------------
@@ -168,148 +144,41 @@ class MemoryStore:
         content: str,
         category: str = "general",
         tags: str = "",
-        *,
-        canonical_key: str = "",
-        metadata: dict | None = None,
-        confidence: float | None = None,
-        source_role: str = "",
-        source_turn: int = -1,
-        observed_at: str = "",
-        contradiction_group: str = "",
-        status: str = "active",
     ) -> int:
         """Insert a fact and return its fact_id.
 
-        Exact duplicates are deduplicated by content. Near-duplicates are
-        normalized by canonical_key, with provenance/evidence merged into the
-        existing row. Contradictions sharing the same contradiction_group remain
-        stored as separate rows and are marked inspectably.
+        Deduplicates by content (UNIQUE constraint). On duplicate, returns
+        the existing fact_id without modifying the row. Extracts entities from
+        the content and links them to the fact.
         """
         with self._lock:
             content = content.strip()
             if not content:
                 raise ValueError("content must not be empty")
 
-            metadata = dict(metadata or {})
-            canonical_key = canonical_key.strip()
-            contradiction_group = contradiction_group.strip()
-            observed_at = observed_at.strip()
-            status = status or "active"
-            trust_score = self.default_trust if confidence is None else _clamp_trust(confidence)
-            metadata_json = json.dumps(metadata, sort_keys=True)
-
-            if canonical_key:
-                existing = self._conn.execute(
-                    "SELECT fact_id, metadata_json, trust_score, confidence, observed_at FROM facts WHERE canonical_key = ?",
-                    (canonical_key,),
-                ).fetchone()
-                if existing is not None:
-                    merged_metadata = self._merge_metadata(existing["metadata_json"], metadata)
-                    merged_trust = max(float(existing["trust_score"]), trust_score)
-                    merged_observed_at = existing["observed_at"] or observed_at
-                    if observed_at and merged_observed_at:
-                        merged_observed_at = min(merged_observed_at, observed_at)
-                    elif observed_at:
-                        merged_observed_at = observed_at
-                    self._conn.execute(
-                        """
-                        UPDATE facts
-                        SET metadata_json = ?,
-                            trust_score = ?,
-                            confidence = ?,
-                            observed_at = ?,
-                            updated_at = CURRENT_TIMESTAMP
-                        WHERE fact_id = ?
-                        """,
-                        (
-                            json.dumps(merged_metadata, sort_keys=True),
-                            merged_trust,
-                            max(float(existing["confidence"] or 0.0), confidence or trust_score),
-                            merged_observed_at,
-                            existing["fact_id"],
-                        ),
-                    )
-                    self._conn.commit()
-                    return int(existing["fact_id"])
-
-            contradiction_rows = []
-            if contradiction_group:
-                contradiction_rows = self._conn.execute(
-                    """
-                    SELECT fact_id, canonical_key, metadata_json
-                    FROM facts
-                    WHERE contradiction_group = ?
-                      AND canonical_key != ?
-                    """,
-                    (contradiction_group, canonical_key),
-                ).fetchall()
-                if contradiction_rows:
-                    status = "contradiction"
-                    metadata = dict(metadata)
-                    metadata["status"] = "contradiction"
-                    metadata["contradiction_group"] = contradiction_group
-                    metadata["contradiction_keys"] = sorted(
-                        {
-                            canonical_key,
-                            *[str(row["canonical_key"]) for row in contradiction_rows if row["canonical_key"]],
-                        }
-                        - {""}
-                    )
-                    metadata_json = json.dumps(metadata, sort_keys=True)
-
             try:
                 cur = self._conn.execute(
                     """
-                    INSERT INTO facts (
-                        content,
-                        category,
-                        tags,
-                        trust_score,
-                        canonical_key,
-                        metadata_json,
-                        confidence,
-                        source_role,
-                        source_turn,
-                        observed_at,
-                        contradiction_group,
-                        status
-                    )
-                    VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)
+                    INSERT INTO facts (content, category, tags, trust_score)
+                    VALUES (?, ?, ?, ?)
                     """,
-                    (
-                        content,
-                        category,
-                        tags,
-                        trust_score,
-                        canonical_key,
-                        metadata_json,
-                        confidence if confidence is not None else trust_score,
-                        source_role,
-                        source_turn,
-                        observed_at,
-                        contradiction_group,
-                        status,
-                    ),
+                    (content, category, tags, self.default_trust),
                 )
                 self._conn.commit()
                 fact_id: int = cur.lastrowid  # type: ignore[assignment]
             except sqlite3.IntegrityError:
+                # Duplicate content — return existing id
                 row = self._conn.execute(
                     "SELECT fact_id FROM facts WHERE content = ?", (content,)
                 ).fetchone()
                 return int(row["fact_id"])
 
-            if contradiction_rows:
-                self._mark_contradictions(
-                    contradiction_group=contradiction_group,
-                    new_canonical_key=canonical_key,
-                    existing_rows=contradiction_rows,
-                )
-
+            # Entity extraction and linking
             for name in self._extract_entities(content):
                 entity_id = self._resolve_entity(name)
                 self._link_fact_entity(fact_id, entity_id)
 
+            # Compute HRR vector after entity linking
             self._compute_hrr_vector(fact_id, content)
             self._rebuild_bank(category)
 
@@ -342,9 +211,6 @@ class MemoryStore:
             sql = f"""
                 SELECT f.fact_id, f.content, f.category, f.tags,
                        f.trust_score, f.retrieval_count, f.helpful_count,
-                       f.canonical_key, f.metadata_json, f.confidence,
-                       f.source_role, f.source_turn, f.observed_at,
-                       f.contradiction_group, f.status,
                        f.created_at, f.updated_at
                 FROM facts f
                 JOIN facts_fts fts ON fts.rowid = f.fact_id
@@ -470,11 +336,7 @@ class MemoryStore:
 
             sql = f"""
                 SELECT fact_id, content, category, tags, trust_score,
-                       retrieval_count, helpful_count,
-                       canonical_key, metadata_json, confidence,
-                       source_role, source_turn, observed_at,
-                       contradiction_group, status,
-                       created_at, updated_at
+                       retrieval_count, helpful_count, created_at, updated_at
                 FROM facts
                 WHERE trust_score >= ?
                   {category_clause}
@@ -525,89 +387,6 @@ class MemoryStore:
                 "helpful_count": row["helpful_count"] + helpful_increment,
             }
 
-    # ------------------------------------------------------------------
-    # Metadata helpers
-    # ------------------------------------------------------------------
-
-    def _load_metadata(self, metadata_json: str | None) -> dict:
-        if not metadata_json:
-            return {}
-        try:
-            data = json.loads(metadata_json)
-            return data if isinstance(data, dict) else {}
-        except Exception:
-            return {}
-
-    def _merge_metadata(self, existing_json: str | None, incoming: dict | None) -> dict:
-        existing = self._load_metadata(existing_json)
-        incoming = dict(incoming or {})
-        merged = dict(existing)
-        merged.update({k: v for k, v in incoming.items() if k not in {"provenance", "evidence", "observation_count", "duplicate_count", "contradiction_keys"}})
-
-        provenance = []
-        seen_provenance: set[str] = set()
-        for item in list(existing.get("provenance", [])) + list(incoming.get("provenance", [])):
-            if item and item not in seen_provenance:
-                seen_provenance.add(item)
-                provenance.append(item)
-
-        evidence = []
-        seen_evidence: set[tuple[str, str, str]] = set()
-        for item in list(existing.get("evidence", [])) + list(incoming.get("evidence", [])):
-            if not isinstance(item, dict):
-                continue
-            key = (
-                str(item.get("provenance", "")),
-                str(item.get("observed_at", "")),
-                str(item.get("source_text", "")),
-            )
-            if key in seen_evidence:
-                continue
-            seen_evidence.add(key)
-            evidence.append(dict(item))
-
-        observation_count = int(existing.get("observation_count", max(1, len(existing.get("evidence", [])) or 1)))
-        observation_count += int(incoming.get("observation_count", max(1, len(incoming.get("evidence", [])) or 1)))
-
-        contradiction_keys = []
-        seen_keys: set[str] = set()
-        for item in list(existing.get("contradiction_keys", [])) + list(incoming.get("contradiction_keys", [])):
-            if item and item not in seen_keys:
-                seen_keys.add(item)
-                contradiction_keys.append(item)
-
-        merged["provenance"] = provenance
-        merged["evidence"] = evidence
-        merged["observation_count"] = observation_count
-        merged["duplicate_count"] = max(0, observation_count - 1)
-        if contradiction_keys:
-            merged["contradiction_keys"] = contradiction_keys
-        return merged
-
-    def _mark_contradictions(self, contradiction_group: str, new_canonical_key: str, existing_rows: list[sqlite3.Row]) -> None:
-        for row in existing_rows:
-            metadata = self._load_metadata(row["metadata_json"])
-            keys = []
-            seen: set[str] = set()
-            for item in list(metadata.get("contradiction_keys", [])) + [new_canonical_key]:
-                if item and item not in seen:
-                    seen.add(item)
-                    keys.append(item)
-            metadata["status"] = "contradiction"
-            metadata["contradiction_group"] = contradiction_group
-            metadata["contradiction_keys"] = keys
-            self._conn.execute(
-                """
-                UPDATE facts
-                SET status = 'contradiction',
-                    metadata_json = ?,
-                    updated_at = CURRENT_TIMESTAMP
-                WHERE fact_id = ?
-                """,
-                (json.dumps(metadata, sort_keys=True), row["fact_id"]),
-            )
-        self._conn.commit()
-
     # ------------------------------------------------------------------
     # Entity helpers
     # ------------------------------------------------------------------
@@ -781,14 +560,8 @@ class MemoryStore:
     # ------------------------------------------------------------------
 
     def _row_to_dict(self, row: sqlite3.Row) -> dict:
-        """Convert a sqlite3.Row to a plain dict with decoded metadata."""
-        data = dict(row)
-        metadata = self._load_metadata(data.get("metadata_json"))
-        if metadata:
-            data["metadata"] = metadata
-            data.setdefault("relation", metadata.get("relation"))
-        data.pop("metadata_json", None)
-        return data
+        """Convert a sqlite3.Row to a plain dict."""
+        return dict(row)
 
     def close(self) -> None:
         """Close the database connection."""

tests/fixtures/memory_extraction_fragments.json

@@ -1,63 +0,0 @@
-{
-  "preferences_and_duplicates": [
-    {
-      "role": "user",
-      "content": "Deploy via Ansible for production changes.",
-      "created_at": "2026-04-22T10:00:00Z"
-    },
-    {
-      "role": "user",
-      "content": "We deploy through Ansible on this repo.",
-      "created_at": "2026-04-22T10:01:00Z"
-    },
-    {
-      "role": "user",
-      "content": "Gitea-first for repository work.",
-      "created_at": "2026-04-22T10:02:00Z"
-    }
-  ],
-  "operational_and_contradictions": [
-    {
-      "role": "user",
-      "content": "The BURN watchdog caps dispatches per cycle to 6.",
-      "created_at": "2026-04-22T11:00:00Z"
-    },
-    {
-      "role": "user",
-      "content": "The provider should stay openai-codex/gpt-5.4.",
-      "created_at": "2026-04-22T11:01:00Z"
-    },
-    {
-      "role": "user",
-      "content": "Correction: the provider should stay mimo-v2-pro.",
-      "created_at": "2026-04-22T11:02:00Z"
-    }
-  ],
-  "mixed_transcript": [
-    {
-      "role": "user",
-      "content": "Deploy via Ansible for production changes.",
-      "created_at": "2026-04-22T10:00:00Z"
-    },
-    {
-      "role": "user",
-      "content": "We deploy through Ansible on this repo.",
-      "created_at": "2026-04-22T10:01:00Z"
-    },
-    {
-      "role": "user",
-      "content": "The BURN watchdog caps dispatches per cycle to 6.",
-      "created_at": "2026-04-22T11:00:00Z"
-    },
-    {
-      "role": "user",
-      "content": "The provider should stay openai-codex/gpt-5.4.",
-      "created_at": "2026-04-22T11:01:00Z"
-    },
-    {
-      "role": "user",
-      "content": "Correction: the provider should stay mimo-v2-pro.",
-      "created_at": "2026-04-22T11:02:00Z"
-    }
-  ]
-}

tests/plugins/memory/test_holographic_auto_extract.py

@@ -1,50 +0,0 @@
-"""Integration tests for holographic auto-extraction with structured fact persistence."""
-
-import json
-import sys
-from pathlib import Path
-
-sys.path.insert(0, str(Path(__file__).resolve().parents[3]))
-
-from plugins.memory.holographic import HolographicMemoryProvider
-
-_FIXTURE_PATH = Path(__file__).resolve().parents[2] / "fixtures" / "memory_extraction_fragments.json"
-
-
-def _load_fixture(name: str):
-    return json.loads(_FIXTURE_PATH.read_text())[name]
-
-
-class TestHolographicAutoExtract:
-    def test_auto_extract_persists_structured_metadata_and_normalizes_duplicates(self, tmp_path):
-        provider = HolographicMemoryProvider(
-            config={
-                "db_path": str(tmp_path / "memory_store.db"),
-                "auto_extract": True,
-                "default_trust": 0.5,
-            }
-        )
-        provider.initialize("test-session")
-
-        messages = _load_fixture("mixed_transcript")
-        provider.on_session_end(messages)
-        provider.on_session_end(messages)
-
-        facts = provider._store.list_facts(min_trust=0.0, limit=20)
-        deploy_facts = [f for f in facts if f.get("relation") == "workflow.deploy_method"]
-        provider_facts = [f for f in facts if f.get("contradiction_group") == "config.provider"]
-
-        assert len(deploy_facts) == 1
-        assert deploy_facts[0]["metadata"]["duplicate_count"] >= 3
-        assert deploy_facts[0]["observed_at"] == "2026-04-22T10:00:00Z"
-        assert deploy_facts[0]["metadata"]["provenance"] == [
-            "conversation:user:0",
-            "conversation:user:1",
-        ]
-
-        assert len(provider_facts) == 2
-        assert {f["status"] for f in provider_facts} == {"contradiction"}
-        assert {f["metadata"]["value"] for f in provider_facts} == {
-            "openai-codex/gpt-5.4",
-            "mimo-v2-pro",
-        }

tests/test_session_compactor.py

@@ -1,6 +1,6 @@
 """Tests for session compaction with fact extraction."""
 
-import json
+import pytest
 import sys
 from pathlib import Path
 
@@ -8,19 +8,12 @@ sys.path.insert(0, str(Path(__file__).resolve().parent.parent))
 
 from agent.session_compactor import (
     ExtractedFact,
-    evaluate_extraction_quality,
-    extract_and_save_facts,
     extract_facts_from_messages,
-    format_facts_summary,
     save_facts_to_store,
+    extract_and_save_facts,
+    format_facts_summary,
 )
 
-_FIXTURE_PATH = Path(__file__).resolve().parent / "fixtures" / "memory_extraction_fragments.json"
-
-
-def _load_fixture(name: str):
-    return json.loads(_FIXTURE_PATH.read_text())[name]
-
 
 class TestFactExtraction:
     def test_extract_preference(self):
@@ -67,48 +60,14 @@ class TestFactExtraction:
             {"role": "user", "content": "I prefer Python."},
         ]
         facts = extract_facts_from_messages(messages)
+        # Should deduplicate
         python_facts = [f for f in facts if "Python" in f.content]
         assert len(python_facts) == 1
 
-    def test_structured_fact_preserves_provenance_and_temporal_metadata(self):
-        facts = extract_facts_from_messages(_load_fixture("preferences_and_duplicates"))
-        deploy_fact = next(f for f in facts if f.relation == "workflow.deploy_method")
-        assert deploy_fact.source_role == "user"
-        assert deploy_fact.source_turn == 0
-        assert deploy_fact.observed_at == "2026-04-22T10:00:00Z"
-        assert deploy_fact.provenance == "conversation:user:0"
-        assert deploy_fact.canonical_key
-        assert deploy_fact.evidence
-        assert deploy_fact.evidence[0]["source_text"].startswith("Deploy via Ansible")
-
-    def test_near_duplicate_facts_are_normalized_into_one_canonical_fact(self):
-        facts = extract_facts_from_messages(_load_fixture("preferences_and_duplicates"))
-        deploy_facts = [f for f in facts if f.relation == "workflow.deploy_method"]
-        assert len(deploy_facts) == 1
-        assert len(deploy_facts[0].evidence) == 2
-        assert deploy_facts[0].metadata["duplicate_count"] == 1
-
-    def test_contradictory_facts_are_preserved_for_unique_slots(self):
-        facts = extract_facts_from_messages(_load_fixture("operational_and_contradictions"))
-        provider_facts = [f for f in facts if f.contradiction_group == "config.provider"]
-        assert len(provider_facts) == 2
-        assert {f.status for f in provider_facts} == {"contradiction"}
-        assert {f.normalized_content for f in provider_facts} == {
-            "openai codex gpt 5 4",
-            "mimo v2 pro",
-        }
-
-    def test_quality_evaluation_reports_noise_reduction(self):
-        metrics = evaluate_extraction_quality(_load_fixture("mixed_transcript"))
-        assert metrics["raw_candidates"] > metrics["normalized_facts"]
-        assert metrics["noise_reduction"] > 0
-        assert metrics["contradiction_groups"] == 1
-
 
 class TestSaveFacts:
     def test_save_with_callback(self):
         saved = []
-
         def mock_save(category, entity, content, trust):
             saved.append({"category": category, "content": content})
 
@@ -117,38 +76,6 @@ class TestSaveFacts:
         assert count == 1
         assert len(saved) == 1
 
-    def test_save_with_extended_callback_metadata(self):
-        saved = []
-
-        def mock_save(category, entity, content, trust, **kwargs):
-            saved.append({
-                "category": category,
-                "entity": entity,
-                "content": content,
-                "trust": trust,
-                **kwargs,
-            })
-
-        fact = ExtractedFact(
-            "project.operational",
-            "watchdog",
-            "BURN watchdog caps dispatches per cycle to 6",
-            0.9,
-            2,
-            source_role="user",
-            observed_at="2026-04-22T11:00:00Z",
-            provenance="conversation:user:2",
-            canonical_key="project.operational|watchdog|dispatch_cap|6",
-            relation="fleet.dispatch_cap",
-            contradiction_group="fleet.dispatch_cap",
-            metadata={"duplicate_count": 0},
-        )
-        count = save_facts_to_store([fact], fact_store_fn=mock_save)
-        assert count == 1
-        assert saved[0]["canonical_key"] == fact.canonical_key
-        assert saved[0]["observed_at"] == "2026-04-22T11:00:00Z"
-        assert saved[0]["metadata"]["duplicate_count"] == 0
-
 
 class TestFormatSummary:
     def test_empty(self):

tests/test_tool_call_benchmark.py

@@ -0,0 +1,115 @@
+"""Tests for Issue #796 tool-calling benchmark coverage and reporting."""
+
+import sys
+from pathlib import Path
+from types import SimpleNamespace
+from unittest.mock import patch
+
+sys.path.insert(0, str(Path(__file__).parent.parent / "benchmarks"))
+
+from tool_call_benchmark import (  # noqa: E402
+    CallResult,
+    DEFAULT_COMPARE_MODELS,
+    ISSUE_796_CATEGORY_COUNTS,
+    ToolCall,
+    generate_report,
+    run_single_test,
+    suite_category_counts,
+)
+
+
+def test_suite_counts_match_issue_796_distribution():
+    counts = suite_category_counts()
+    assert counts == ISSUE_796_CATEGORY_COUNTS
+    assert sum(counts.values()) == 100
+
+
+def test_default_compare_models_cover_issue_796_lanes():
+    assert len(DEFAULT_COMPARE_MODELS) == 3
+    assert any("gemma-4-31b" in spec for spec in DEFAULT_COMPARE_MODELS)
+    assert any("gemma-4-26b" in spec for spec in DEFAULT_COMPARE_MODELS)
+    assert any("mimo-v2-pro" in spec for spec in DEFAULT_COMPARE_MODELS)
+
+
+def test_generate_report_includes_parallel_and_cost_metrics(tmp_path):
+    output_path = tmp_path / "report.md"
+    results = [
+        CallResult(
+            test_id="file-01",
+            category="file",
+            model="gemma-4-31b",
+            prompt="Read the file.",
+            expected_tool="read_file",
+            success=True,
+            tool_called="read_file",
+            schema_ok=True,
+            tool_args_valid=True,
+            execution_ok=True,
+            tool_count=2,
+            parallel_ok=True,
+            latency_s=1.25,
+            total_tokens=123,
+            estimated_cost_usd=0.0012,
+            cost_status="estimated",
+        ),
+        CallResult(
+            test_id="web-01",
+            category="web",
+            model="mimo-v2-pro",
+            prompt="Search the web.",
+            expected_tool="web_search",
+            success=False,
+            tool_called="web_search",
+            schema_ok=True,
+            tool_args_valid=False,
+            execution_ok=False,
+            tool_count=1,
+            parallel_ok=False,
+            latency_s=2.5,
+            error="bad args",
+            total_tokens=456,
+            estimated_cost_usd=None,
+            cost_status="unknown",
+            skipped=True,
+            skip_reason="web_search unavailable",
+        ),
+    ]
+
+    report = generate_report(results, ["gemma-4-31b", "mimo-v2-pro"], output_path)
+
+    assert output_path.exists()
+    assert "Parallel tool success" in report
+    assert "Avg token cost per call (USD)" in report
+    assert "Skipped / unavailable" in report
+    assert "Requested category mix" in report
+
+
+def test_run_single_test_skips_when_expected_tool_unavailable():
+    class FakeAgent:
+        def __init__(self, *args, **kwargs):
+            self.valid_tool_names = {"read_file", "terminal"}
+            self.session_input_tokens = 0
+            self.session_output_tokens = 0
+            self.session_cache_read_tokens = 0
+            self.session_cache_write_tokens = 0
+            self.session_api_calls = 0
+            self.base_url = ""
+            self.api_key = None
+
+        def run_conversation(self, *args, **kwargs):
+            raise AssertionError("run_conversation should not be called for unavailable tools")
+
+    tc = ToolCall(
+        id="mcp-01",
+        category="mcp",
+        prompt="Use an MCP tool to list resources.",
+        expected_tool="",
+        expected_tool_prefix="mcp_",
+    )
+
+    with patch.dict(sys.modules, {"run_agent": SimpleNamespace(AIAgent=FakeAgent)}):
+        result = run_single_test(tc, "gemini:gemma-4-31b-it", "gemini")
+
+    assert result.skipped is True
+    assert "mcp_" in result.skip_reason
+    assert result.success is False